Objective This study aims to construct a site quality evaluation model for uneven-aged natural Picea schrenkiana forests in the Tianshan Mountains, based on the site form method and the height-DBH relationship. It also seeks to verify the model’s applicability in the absence of stand age data, thereby providing technical support for sustainable forest management and productivity prediction.

Method Field-measured diameter at breast height (DBH) data were integrated with tree height data derived from airborne LiDAR point clouds. Five dominant-tree combinations were used to calculate dominant height and quadratic mean diameter, and the optimal guide curve was selected accordingly. The reference diameter was determined based on growth characteristics, and site classes were classified using the site form method. Site class was then incorporated as a random effect into a height-DBH nonlinear mixed-effects model, and model performance was comprehensively evaluated using multiple statistical criteria.

Result (1) The Schumacher guide curve fitted using the H₅-D_g,5 combination, derived from the adjusted largest trees method with DBH-based ranking, achieved the best performance (R² = 0.794 3, RMSE = 1.309 3); (2) The reference diameter was determined to be 34 cm. On this basis, the exponential height-DBH model incorporating a shifted reciprocal term of DBH was identified as the optimal base model, with R² = 0.833 9 and RMSE = 2.984 3 for the modeling dataset, and R² =0.858 5 and RMSE = 2.822 3 for the validation dataset; (3) The introduction of the nonlinear mixed-effects model improved model accuracy. The H₂-D_g,2 combination, derived from the conventional estimation method with tree-height-based ranking, performed best during the modeling stage (R² =0.845 8, RMSE = 2.875 6), whereas the H₆-D_g,6 combination, derived from the adjusted largest trees method with tree-height-based ranking, showed the best performance during the validation stage (R² = 0.875 6, RMSE = 2.645 5).

Conclusion The combination of the site form method and height-DBH models enables reliable site quality classification for uneven-aged natural Picea schrenkiana forests in the absence of age data. The selection of dominant tree combinations significantly influences guide curve construction and model performance. Incorporating site class as a random effect effectively improves the fitting accuracy and predictive ability of height-DBH models, demonstrating that this approach provides a practical and reliable technique for site quality evaluation in uneven-aged forests.